Recovery from Alcohol Problemswith and without Treatment:Prevalence in Two Population Surveys

Linda C. Sobell, PhD, John A. Cunningham, PhD, and Mark B. Sobell, PhD

IntroductionEven though estimates of the ratio of

untreated to treated individuals withalcohol problems in the general popula-tion range from 3:1 to 13:1,1 there hasbeen little interest in the fates of un-treated individuals, including whetherthey can recover without treatment (natu-ral recovery).2- Recently, prominent orga-nizations such as the Institute of Medi-cine5 and the American PsychiatricAssociation6 have acknowledged that natu-ral recoveries constitute a significantpathway to recovery from alcohol prob-lems. However, while epidemiologicaland longitudinal studies of alcohol prob-lems in the general population have led tospeculation about the prevalence of un-treated recoveries, these studies have notdirectly assessed recovery rates for bothtreated and untreated individuals orwhether drinking outcomes involved absti-nence or moderate drinking." 4'7-9

Traditionally, alcohol problems havebeen viewed from the context of individu-als who are severely dependent on alco-hol.10 However, epidemiological studiesshow that while persons with severealcohol problems constitute the majorityof individuals in treatment programs, theyrepresent a minority of those with alcoholproblems.1011 It has been estimated thatthe ratio of problem drinkers (i.e., mildalcohol dependence) to those severelydependent on alcohol is about 4:1.5Although severely dependent alcoholabusers have more serious problems, mostalcohol-related costs to society stem fromthe large numbers of problem drinkers(e.g., drunk driving, days of missed work,domestic violence).5'12 From a publichealth perspective, the fate of all individu-als with alcohol problems is important.

In two recent surveys, randomlyselected adults in the general population

were asked a broad range of questionsabout their past and present use ofalcohol. Using data from these surveys,this report examines the prevalence oftreated and untreated recoveries fromalcohol problems. Because research indi-cates that individuals with low-severityproblems often recover by reducing ratherthan stopping drinking, recoveries werealso classified as involving abstinence ormoderate drinking.

MethodsData were derived from two surveys:

the National Alcohol and Drugs Survey("National Survey"), conducted by Statis-tics Canada in March 1989,13 and theOntario Alcohol and Drug Opinion Sur-vey ("Ontario Survey"), conducted by theInstitute of Social Research at YorkUniversity (Toronto) in April and May1993.14 In both surveys, persons living ininstitutions (e.g., those in hospitals) wereexcluded. All respondents were assuredof anonymity (i.e., names were neverasked). Random-digit dialing was thesampling method used in both surveys.14"5Households without telephones (<2%)could not be covered by this samplingmethod.

Linda C. Sobell is with the Addiction ResearchFoundation and the Departments of Psychol-ogy, Family and Community Medicine, andBehavioural Science, University of Toronto,Toronto, Ontario, Canada. John A. Cunning-ham is with the Addiction Research Founda-tion and the Department of Psychology, Univer-sity of Toronto. Mark B. Sobell is with NovaSoutheastern University, Ft. Lauderdale, Fla.

Requests for reprints should be sent toLinda C. Sobell, PhD, Center for PsychologicalStudies, Nova Southeastern University, 3301College Ave, Ft. Lauderdale, FL 33314.

This paper was accepted February 1,1996.

July 1996, Vol. 86, No. 7

Natural Recovery

In the National Survey, telephoneinterviews were conducted with 11 634Canadians, 15 years of age or older, fromall 10 provinces. Persons living in theYukon and Northwest territories were notinterviewed because a sizable percentagedid not have telephones. Interviewersrandomly selected one potential respon-dent from within each household tocomplete the questionnaire (no proxy orsubstitution interviews were allowed).The overall response rate was 78.7%.Reasons for nonresponses included thefollowing: refusal, illness, absent duringsurvey, language problems, and no con-tact. The final sample was weighted toensure representativeness and to compen-sate for survey nonresponse (e.g., house-holds without telephones, households withmultiple telephones, number of respon-dents within households, and census pro-jection counts for each province). Theexact weighting procedure and samplingdesign have been discussed elsewhere.15The survey period was chosen to avoidholiday periods, which might overrepre-sent drinking by respondents (this proce-dure was not used in the Ontario Sur-vey).'5

In the Ontario Survey, telephoneinterviews were conducted with 1034respondents 18 years of age or older. Onerespondent from within each householdwas randomly selected to complete theinterview. Respondents had to speak oneof Canada's two official languages (En-glish or French; only six interviews wereconducted in French). As a means ofensuring a representative sample andcompensating for potential sources ofsurvey error, the final overall responserate of 65% was weighted by consideringthe number of telephones and adults ineach household to be a probability sampleof adults in Ontario. An average designweight for a selection of variables wascalculated, and the sample was down-weighted to all cases to take account ofdesign effects in tests of significance. Theweighting procedure and sampling designfor this survey have been described else-where.'4

Of the original National Survey andOntario Survey samples, data from respon-dents 20 years of age or older (in eachsurvey, this was the closest age category tothe legal drinking age of 19 in Canada)were considered for the present study(n = 10 796 and n = 1001, respectively).For all analyses, unweighted values areused in reporting sample sizes, whileresults from statistical tests are based onweighted data.

Both surveys asked all respondentsquestions that allowed for a determina-tion of past or current alcohol problems(i.e., frequency and quantity, problemsassociated with use). (A copy of thequestions used in both surveys is availablefrom the first author on request.) Ifrespondents had had an alcohol problemand had resolved it, they were askedquestions about their drinking that al-lowed for a determination as to whethertheir resolution was to abstinence ornonabstinence (i.e., moderate drinking).They were also asked whether they hadever received any formal services, help, ortreatment for an alcohol-related problem.In the National Survey, respondents werealso asked the approximate length of theirresolution. Several background questions(e.g., age, education) were also asked ofall survey respondents.

The drinking criteria chosen for menand women in the present study wereconsistent with definitions used in previ-ous research.1620 The sample size andcriteria for classifying subjects into each ofthe four groups are noted subsequently.Lifetime abstainers and respondents notmeeting any of the following criteria wereexcluded from subsequent analyses.

Respondents who were classified as"resolved abstinent" (National Survey,n = 302; Ontario Survey, n = 34) werecurrent abstainers who reported pastproblems related to their alcohol use andhad quit drinking for at least 1 year priorto the interview. These respondents hadto report experiencing at least one of thefollowing types of alcohol problems priorto their resolution: (1) problems affectingtheir work, studies, or employment oppor-tunities; (2) problems interfering withtheir family or home life; (3) problemsaffecting their physical health; (4) prob-lems affecting their friendships or sociallife; and (5) problems affecting theirfinancial position.

Those who were classified as "re-solved nonabstinent" (National Survey,n = 144; Ontario Survey, n = 60) werecurrent drinkers who reported past prob-lems related to their alcohol use (sameproblems as for resolved abstinent sub-jects) and had reduced their drinking to anonproblem level for at least 1 year priorto the interview. These individuals re-ported experiencing no current (i.e., past12 months) problems due to their alcoholconsumption, and their drinking levels inthe past 12 months (i.e., the periodcovered by the survey) were not consid-ered to constitute gender-related healthrisks.'6'8 All respondents were first asked,

"During the last 12 months, how often onaverage did you drink alcoholic bever-ages?" They were then asked, "On thedays when you drank, how many drinksdid you usually have?" When respondentswere asked about drinks, they were toldthat the word "drink" meant "one straightor mixed drink with one ounce and a halfof hard liquor," "one bottle of beer orglass of draft," or "one glass of wine or awine cooler." Allowable drinking wasdefined as (1) usual drinking of threedrinks or fewer for men and two drinks orfewer for women; (2) no more than 2 daysof five to seven drinks in the past year (toallow for a small amount of celebratorydrinking [e.g., holidays]); and (3) a maxi-mum number of drinks consumed on anyone occasion in the past year of seven.

Those who were classified as currentproblem drinkers (National Survey,n = 1158; Ontario Survey, n = 104) werecurrent drinkers who reported experienc-ing problems due to their alcohol use inthe previous year (same problem defini-tion as for resolved abstinent respon-dents) or who drank at a level associatedwith health risks. For men, the latter wasdefined as usually drinking seven drinksor more on days when they drank; forwomen, it was defined as usually drinkingfive drinks or more.19'20 Respondents alsohad to report drinking at least one timeper week in the past year.

Those who were classified as "cur-rent social drinkers" (National Survey,n = 3319; Ontario Survey, n = 405) werecurrent drinkers who reported no priorproblems associated with their alcohol useand who drank at levels that would not beconsidered as presenting a potential healthrisk (see definition for resolved nonabsti-nent respondents).

All resolved respondents in bothsurveys were further classified as towhether they had ever used any of thefollowing types of treatment, help, orservices related to their alcohol problem:Alcoholics Anonymous or other supportgroup; psychologist, psychiatrist, or socialworker; psychiatric hospital; minister,priest, or rabbi; doctor or nurse; hospitalor emergency department; alcohol/drugaddiction agency; or detoxification centeror halfway house. Ontario Survey respon-dents were also asked whether they hadbeen in a drinking and driving programbecause of their drinking, but there wereno affirmative responses among the re-solved respondents.

American Journal of Public Health 967July 1996, Vol. 86, No. 7

National Survey (N = 446) Ontario Survey (N = 94)

Treatment

No Treatment No Treatment

National Survey (N = 302)

_,9jjjMjbh, Treatment

No Treatment

Ontario Survey (N = 34)

_4~; Treatment

No Treatment

National Survey (N = 144)Treatment

(n.=9)

No Treatment

Ontario Survey (N = 60)

9.4%(n= 7)

No

FIGURE 1-Prevalence rates of abstinence and nonabstinence recoveries forIndividuals who had recovered (for 1 year or more) from an alcoholproblem with and without treatment: the Canadian 1989 NationalAlcohol and Drug Survey and the 1993 Ontario Alcohol and DrugOpinion Survey.

ResultsFigure 1 shows that of all respon-

dents in both surveys who reported resolv-ing an alcohol problem, 77.5% (n = 322;National Survey) and 77.7% (n = 70;Ontario Survey) did so without formaltreatment or help. The remaining 22.5%(n = 124) and 22.3% (n = 24) in eachsurvey reported using some type of alco-hol-related service, self-help group, or

counseling. Furthermore, of all recoveriesin the National and Ontario surveys,62.0% (n = 302) and 37.3% (n = 34) of

respondents, respectively, reported return-ing to abstinence; the remaining recover-

ies, 38.0% (n = 144) and 62.7% (n = 53),respectively, involved moderation out-comes. More strikingly, as shown inFigure 1, in both surveys almost all (96.7%and 90.6%, respectively) of the moderatedrinking recoveries involved respondentswho had never received any alcohol-related help or treatment. The majority(65.7% and 56.0%, respectively) of absti-nent recoveries also occurred in theabsence of treatment. In both surveys, ofall respondents who said they had used

some type of help or treatment for theiralcohol problem, the predominant re-

source reported was Alcoholics Anony-mous (National Survey, 82.8%; OntarioSurvey, 52.8%).

When treatment use by drinkingrecovery type was examined, chi-squaretests revealed significant (both Ps < .001)> differences for each survey, suggestingthat nonabstinent recoveries were more

likely to occur among respondents whorecovered without treatment than amongtreated individuals. Because resolved non-abstinent-treatment recoveries consti-tuted a very small number of cases in bothsurveys (National Survey: 3.3%, n = 9;Ontario Survey: 9.4%, n = 7), they were

excluded from further analyses.Table 1 presents comparisons of

sociodemographic and alcohol-relatedvariables across the three resolved groups

in both surveys: resolved abstinent-treatment, resolved abstinent-no treat-ment, and resolved nonabstinent-no treat-ment. Chi-square tests were performedon nominal (nonparametric) variables,and one-way analyses of variance(ANOVAs) were performed on quantita-tive (parametric) variables.

In the National Survey, as shown inTable 1, there were significant differencesbetween the three groups for 7 of the 10variables. Respondents who returned tomoderate drinking, in comparison withthose who were abstinent, were more

likely to be female, to be younger, to havesome postsecondary education, to havehigher incomes, and to have stoppedsmoking (among those who ever smoked;nondaily smokers were excluded), andfewer had two or more alcohol problemsprior to their resolution. In the OntarioSurvey, with two exceptions, there were

no significant (P > .05) differences be-tween the three groups on variables forwhich chi-square tests could be per-

formed (i.e., for some variables, chi-square tests could not be performedbecause more than 20% of cells hadexpected frequencies of less than 5). As inthe National Survey, a greater percentageof Ontario Survey respondents in treat-ment reported two or more alcohol-related problems. In both surveys, one-

way ANOVAs showed that respondentsin treatment reported a significantlygreater mean number of alcohol-relatedproblems. For the Ontario Survey, Scheffepost hoc comparisons revealed that theresolved abstinent-treatment group dif-fered significantly from the other two

July 1996, Vol. 86, No.7968 American Journal of Public Health

Sobell et al.

..

Natural Recovery

groups. For the National Survey, Scheffepost hoc tests revealed that all groupsdiffered significantly, with the resolvedabstinent-treatment respondents report-ing the greatest number of alcohol-relatedproblems prior to their resolution and theresolved nonabstinent-no treatment re-

spondents the fewest.For those individuals with alcohol

problems who reported returning to mod-erate drinking, a major issue is whethertheir postrecovery drinking was like thatof social drinkers in the general popula-tion. To address this, in both surveys theresolved nonabstinent-no treatment re-

spondents' drinking in the past year was

compared with the drinking of the othertwo current drinker groups: social drink-ers and problem drinkers. Table 2 showsthe means and standard deviations for sixdrinking variables for the three drinkergroups in each survey. For two of the sixvariables, Ontario Survey findings couldnot be examined because data parallel tothose of the National Survey were notcollected. One-way ANOVAs were per-formed on all variables for both surveys,

and all of these analyses yielded signifi-cant differences. With one exception, the

comparisons between the three currentdrinker groups in each survey showed thatthe resolved nonabstinent and socialdrinker groups were similar and that bothdiffered significantly from the problemdrinker group (Scheffe post hoc tests).The exception occurred for mean numberof days drinking in the past week, forwhich Scheffe post hoc comparisonsshowed all groups to differ significantly.As shown in Table 2, problem drinkers

reported more mean days drinking in the

past week than the resolved nonabstinent

respondents, who in turn reported more

drinking days than the social drinkers.

DiscussionThree major findings emerged from

the two surveys. First, more than three

quarters of all adults in both surveys whorecovered from an alcohol problem for a

year or more did so without formal help or

treatment. Of note, this finding is consis-

tent with a large body of evidence showing

that of all cigarette smokers, 80% to 90%

stop on their own.21'22 Second, these

remarkably similar findings came from

two independent surveys conducted a few

years apart by two different interview

groups. The fact that the results from the

two surveys parallel each other suggeststhat conclusions based on these findingsare likely to be robust. Third, although all

respondents had to be recovered for a

minimum of 1 year or more, more than

half of all respondents in the National

Survey, in which length of recovery was

assessed, reported having been recovered

for more than 5 years. This finding is

significant because two other studies of

treated23 and untreated24 recovered alco-

hol abusers have shown that relapses are

highly unlikely after 5 years.Although these findings may be seen

as inconsistent with the traditional model

of alcoholism,25 they must be viewed in

the context that "the types of problemsreported in surveys are rarely as severe

as those observed in alcoholism clin-ics."26(p72) From a public health stand-

point, the critical issue related to societal

American Journal of Public Health 969July 1996, Vol. 86, No. 7

TABLE 1 Sociodemographic Characteristics of Three Resolved Drinker Groups from Two Canadian Surveys

Group

Resolved Abstinent Resolved Abstinent Resolved Nonabstinentwith Treatment with No Treatment with No Treatment

pa

National Ontario National Ontario National OntarioSurvey Survey Survey Survey Survey Survey National Ontario

(n = 115) (n = 17) (n = 187) (n = 17) (n = 135) (n = 53) Survey Survey

Male, % (no.) 75.9 (91) 62.3 (10) 79.1 (137) 28.7 (5) 61.6 (85) 47.5 (25) .005 NSAge <34 years, % (no.) 14.0 (19) 15.1 (2) 18.8 (39) 29.7 (5) 36.9 (53) 32.2 (15) <.001 ... bFamily income > $40 000, 21.2 (20) 69.8 (12) 28.6 (38) 77.2 (13) 57.5 (62) 71.6 (37) <.001 ...b% (no.)

Married, % (no.) 68.7 (69) 62.3 (10) 66.9 (115) 67.3 (11) 72.4 (90) 58.5 (27) NS NSSome postsecondary 21.8 (21) 47.2 (9) 19.0 (32) 47.5 (7) 44.9 (54) 55.7 (28) <.001 NS

education, % (no.)White collar,c% (no.) 26.8 (19) 30.6 (6) 23.8 (24) 40.0 (5) 39.3 (36) 43.0 (23) NS NSResolved alcohol problem 58.9 (70) ...d 58.3 (115) ...d 54.4 (72) ...d NS ...d> 5 years, % (no.)

Current ever smokers,e 73.9 (68) 41.0 (5) 57.3 (72) 44.1 (5) 37.2 (34) 28.6 (7) <.001 ...

% (no.)2 or more past alcohol 81.6 (96) 84.9 (15) 48.3 (83) 38.6 (6) 27.9 (35) 38.8 (21) <.001 <.01

problems,f % (no.)No. of past alcohol prob- 3.6 (1.6) 3.6 (1.7) 2.0 (1.3) 1.8 (1.3) 1.5 (1.0) 1.8 (1.2) <.001 h <.001

lems,9 mean (SD)

Note. National Survey = 1989 National Alcohol and Drugs Survey; Ontario Survey = 1993 Ontario Alcohol and Drug Opinion Survey. Percentages, means,and standard deviations were based on weighted values. Sample sizes varied for some of the characteristics measured. NS = nonsignificant (P > .05).

aAll analyses were chi-square tests except for mean number of alcohol consequences (one-way analysis of variance).bNot determinable because more than 20% of cells had expected frequencies of less than 5.CBoth surveys: students and homemakers excluded; National Survey: respondents who worked in past year; Ontario Survey: current or prior employment.dData for this variable not available.eNondaily smokers excluded.The percentages of problem drinkers in the National and Ontario surveys with 2 or more reported alcohol problems in the previous 12 months were 39.8%and 50.7%, respectively.

sOut of a maximum of 5 possible problems.hAll groups differed significantly from each other (Scheffe post hoc comparisons, P < .05).'Resolved abstinent-treatment group differed significantly from the other two groups (Scheffe post hoc comparisons, P < .05).

Sobell et al.

costs of alcohol problems is that problemsof any nature and number enter into thefinal equation. This is important becausepersons whose problems are not severe

account for the preponderance of costs(i.e., alcohol problems "occur at lowerrates but among much greater numbers as

one moves from the heaviest drinkers tomore moderate drinkers"27[P4").

Both surveys revealed that a substan-tial number of all recoveries involvedindividuals who resolved their alcoholproblem by reducing their drinking tolevels that would not be considered a

health risk and did not incur conse-

quences. The current drinking of theresolved respondents who engaged inmoderate drinking resembled that ofmembers of the social drinker group whoreported never having had an alcoholproblem. There was a large differencebetween the two surveys in terms of theprevalence of nonabstinence recoveries(National Survey, 38%; Ontario Survey,63%). However, when abstinence andnonabstinence prevalence rates for On-tario respondents (50.8% and 49.2%,respectively) in the National Survey were

compared with those of Ontario Survey

respondents (37.5% and 62.5%, respec-

tively), a chi-square test revealed no

significance difference (P > .05). Thus,nonabstinence recovery rates appear tobe higher in the province of Ontario thanin the national sample. These figures werenot unexpected; several studies haveshown that sociocultural factors not onlyare related to problem severity but alsoappear to be associated with recovery

type.27-30 Specifically, of those who re-

cover without treatment, resolved nonab-stinent individuals are of higher socioeco-nomic status and have higher incomes andmore education than resolved abstinentindividuals. With respect to the provinceof Ontario, statistics show that individualsin this province have higher incomes andgreater educational levels than do those inmany of the other provinces.'5

The fact that almost all of thenonabstinent recoveries in both surveysoccurred in the absence of treatmentparallels findings from two other studies.In a study involving short-term recoveries(6 months), more untreated (45%) thantreated (26%) alcohol abusers returned tononproblem drinking.27 In a study of onlyuntreated problem drinkers, almost all

recoveries involved a return to moderatedrinking.3' Another finding in both sur-

veys that is consistent with results fromother studies is that a greater percentageofwomen with alcohol problems returnedto moderate drinking than to absti-nence.4,5,30,32

In both surveys, resolved respon-dents who had been in treatment reportedalmost twice as many alcohol-relatedconsequences as respondents who re-

solved without treatment. These data,coupled with the high percentage of suchrespondents reporting more than one

alcohol problem prior to their resolution,suggest that respondents who had been intreatment had more serious alcohol prob-lems than those who recovered on theirown. Similar findings have been reportedfor long-term naturally recovered alcoholabusers recruited by advertisements.30

As with almost all major general-population surveys, the present surveyshad methodological limitations. First, theyinvolved self-reported data, retrospectivereports, and no formal diagnostic assess-

ment (e.g., Diagnostic and StatisticalManual of Mental Disorders [4th edition;DSM-IV]).6 Second, drinking assessments

July 1996, Vol. 86, No. 7970 American Journal of Public Health

TABLE 2-Means and Standard Deviations for Drinking Variables: Three Current Drinker Groups from Two Canadian Surveys

Group

Resolved Nonabstinentwith No Treatment, Social Drinkers, Problem Drinkers,

Mean (SD) Mean (SD) Mean (SD)p

National Ontario National Ontario National OntarioSurvey Survey Survey Survey Survey Survey National Ontario

(n = 135) (n = 53) (n = 3319) (n = 405) (n = 1158) (n = 104) Survey Survey

No. drinks per drinking 1.7 (0.6) 1.6 (0.6) 1.6 (0.6) 1.5 (0.6) 5.5 (4.4) 5.7 (3.6) <.001 a <.001 a

day (average yearlyestimate)

No. drinks per drinking 1.7 (0.7) ... c 1.5 (0.7) ...c 4.7 (4.0) .. c <.001 a c

day in past week (dailycalendar)b

Total no. drinks in 2.4 (3.9) 1.8 (2.5) 1.4 (2.7) 1.9 (3.0) 10.0 (15.5) 13.3 (14.6) <.001a <.001apast week

No. days during past year 0.4 (0.7) 0.1 (0.4) 0.2 (0.6) 0.1 (0.4) 39.6 (69.1) 25.2 (28.1) <.001a <.o1ain which .5 drinks wereconsumed

Greatest no. drinks on 3.4 (1.5) 3.0 (1.2) 2.8 (1.4) 2.7 (1.4) 11.4 (8.7) 13.2 (7.4) <.001 a <.001 a

any one occasion inpast year

No. days drinking in 1.5 (2.3) ... 0.9 (1.7) ... 2.1 (2.0) ... <.0.d..

past week

Note. National Survey = 1989 National Alcohol and Drug Survey; Ontario Survey = 1993 Ontario Alcohol and Drug Opinion Survey. Percentages, means,and standard deviations were derived from unweighted data. Sample sizes varied for some of the variables measured. All analyses were one-way analysesof variance.

aResolved nonabstinent and social drinker groups differed significantly from the problem drinker group (Scheffe post hoc comparisons, P < .05).bOf those subjects who drank in the previous week.cData for this variable were not collected in the Ontario survey.dAII groups differed significantly from each other (Schefft post hoc comparisons, P < .05).

Natural Recovery

involved a short time frame (i.e., pastweek or past year). Third, the anonymitygiven respondents could have encourageddishonesty (i.e., knowing that one's an-swers could not be checked), although itcould as easily have promoted moreaccurate reporting. Finally, both surveysinvolved telephone interviews (note, how-ever, that over the years, survey researchhas increasingly relied on telephone inter-views; in reviewing different interviewmethods, Polich and Kaelber concludedthat there does not seem to be a reason toprefer a certain method over another26).

Although none of the questions inthe two surveys allowed for a determina-tion of the severity of alcohol problems,all past (resolved) and current problemdrinkers did report experiencing somealcohol-related problems. Resolved drink-ers in the two surveys reported an averageof 2.2 and 2.1 (National Survey andOntario Survey, respectively) alcohol prob-lems (out of 5 possible problems) prior totheir recovery, and close to half of therespondents (42%, National Survey; 47%,Ontario Survey) in each survey reported 2or more problems. While some mightargue that experiencing a single alcoholproblem is not sufficient grounds forclassifying one as a problem drinker, thisdefinition is consistent with the criteria foran alcohol abuse diagnosis in the DSM-IV, according to which a maladaptivepattern of use leads to "clinically signifi-cant impairment or distress, as manifestedby one (or more) of the following within a12-month period'"6(P182) (the problemslisted in the DSM-IV are similar to thoseused in the current surveys [i.e., work,school, home, significant social or interper-sonal problems]).

Although there were no externalvalidity checks, several factors add confi-dence to the data. First, although the twosurveys were conducted in different loca-tions, in different years, and by differentorganizations, their results were consis-tent. Support also comes from anotherrecent study of adults who visited ascience center in Ontario.33 Three quar-ters (48/64) of those who had resolved analcohol problem for 1 year or more did sowithout treatment, and 57.8% (37/64)resolved through nonabstinence. Second,in the National Survey, two differentlyworded but similar questions about drink-ing (i.e., respondents' mean estimateddrinks per day in the past year using acalendar and reported mean drinks perdrinking day in the week prior to theinterview) captured almost identical drink-ing patterns for each of the groups of

drinkers (see Table 2). Third, the propor-tions of individuals reporting problemswith alcohol to all identified drinkers inboth surveys (National Survey, 10.3%;Ontario Survey, 11.5%) were similar toprevalence figures in other major popula-tion surveys.26'34,35

Although the results from the pres-ent two surveys were strikingly similar,one notable difference (see Table 1) wasthat the Ontario Survey, in comparisonwith the National Survey, included morerespondents who were younger, werefemale, had some postsecondary educa-tion, had higher incomes, and had white-collar jobs. To investigate whether thesedifferences were related to the affluenceof the province of Ontario as comparedwith other parts of Canada, we examinedNational Survey data for respondentsfrom Ontario separately. Ontario respon-dents in the National Survey were foundto have characteristics similar to those ofrespondents in the Ontario Survey. Thus,it appears that national survey respon-dents differ demographically from thosein a particular province, a finding alsoobserved in surveys between states in theUnited States.36,37 Although demographicdifferences were observed between theNational Survey and the Ontario Survey,the important point is that respondents'drinking and natural recovery rates werevery similar.

ConclusionsData from the two general popula-

tion surveys reported in this paper al-lowed the calculation of prevalence ratesof recoveries from alcohol problems withand without treatment. The findings fromthese two surveys significantly bolster thegrowing body of studies showing thatmany individuals with alcohol problemsrecover on their own. Furthermore, asizable proportion of individuals reporteddrinking in a moderate nonproblem man-ner after resolving their problem. In thisregard, and as noted in another study ofnatural recoveries, it is unclear whetherwe have identified multiple pathways outof the same kind of alcohol problem ordifferent types of alcohol problems.28Answers to this question must awaitfuture longitudinal research.

From a public health perspective, thestudy of individuals with alcohol problemswho never receive treatment is importantbecause most will never come to theattention of clinical service providers andbecause the majority of recoveries fromalcohol problems occur outside of clinical

programs. For these reasons, is importantto broaden our perspective on alcoholproblems from a clinic phenomenon to apublic health problem. O

AcknowledgmentThe research presented in this paper wassupported, in part, by grant AA08593 from theNational Institute on Alcohol Abuse andAlcoholism.

References1. Roizen R. Barriers to Alcoholism Treatment.

Berkeley, Calif: Alcohol Research Group;1977.

2. Smart RG, Schmidt W, Moss MK. Socialclass as a determinant of the type andduration of therapy received by alcoholics.IntJAddict. 1969;4:543-556.

3. Tuchfeld BS. Spontaneous remission inalcoholics: empirical observations and theo-retical implications.JStudAlcohol. 1981;42:626-641.

4. Sobell LC, Sobell MB, Toneatto T. Recov-ery from alcohol problems without treat-ment. In: Heather N, Miller WR, GreeleyJ, eds. Self-Control and the Addictive Behav-iours. New York, NY: Maxwell MacMillan;1992.

5. Institute of Medicine. Broadening the BaseofTreatmentforAlcohol Problems. Washing-ton, DC: National Academy Press; 1990.

6. Diagnostic and Statistical Manual ofMentalDisorders. 4th ed. Washington, DC: Ameri-can Psychiatric Association; 1994.

7. Fillmore KM, Hartka E, Johnstone BM,Leino EV, Motoyoshi M, Temple MT.Preliminary results from a meta-analysis ofdrinking behavior in multiple longitudinalstudies. BrJAddict. 1991;86:1203-1210.

8. Fillmore KM, Hartka E, Johnstone BM,Speiglman R, Temple MT. SpontaneousRemission of Alcohol Problems: A CriticalReview. Washington, DC; Institute of Medi-cine; 1988.

9. Fillmore KM. Alcohol Use Across the LifeCourse: A Critical Review of 70 Years ofIntemational Longitudinal Research.Toronto, Ontario, Canada: Addiction Re-search Foundation; 1988.

10. Cahalan D. Studying drinking problemsrather than alcoholism. In: Galanter M, ed.Recent Developments in Alcoholism. NewYork, NY: Plenum Press; 1987; 5.

11. Room R. Measurement and distribution ofdrinking patterns and problems in generalpopulations. In: Edwards G, Gross MM,Keller M, Moser J, Room R, eds. Alcohol-Related Disabilities. Geneva, Switzerland:World Health Organization; 1977.

12. Moore MH, Gerstein DR. Alcohol andPublic Policy: Beyond the Shadow ofProhibi-tion. Washington, DC: National AcademyPress; 1981.

13. National Alcohol and Other Drugs Survey(1989): A Highlight Report. Ottawa, On-tario, Canada: Health and Welfare Canada;1990.

14. Northrup DA. The 1993 Adult OpinionSurvey: A Survey ofOntario Residents for theAddietion Research Foundation. TechnicalDocumentation. Toronto, Ontario, Canada:Institute for Social Research, York Univer-sity; 1993.

July 1996, Vol. 86, No. 7 American Journal of Public Health 971

Sobell et al.

15. National Alcohol and Other Dnrgs Survey:Microdata Documentation and Users Guide.Ottawa, Ontario, Canada: StatisticsCanada; 1990.

16. Anderson P, Cremona A, Paton A, TurnerC, Wallace P. The risk of alcohol. Addic-tion. 1993;88:1493-1508.

17. Babor TF, Kranzler HR, Lauerman RJ.Social drinking as a health and psychoso-cial risk factor. Anstie's limit revisited. In:Galanter M, ed. Recent Developments inAlcoholism. New York, NY: Plenum Press;1987; 5.

18. Rimm RB, Giovannucci EL, Willett WC,et al. Prospective study of alcohol consump-tion and risk of coronary disease in men.Lancet. 1991;338:464-468.

19. Catarino PA. Is there a safe level ofdrinking? A student's view. Alcohol Alco-hoL 1992;27:465-470.

20. Kranzler HR, Babor TF, Lauerman RJ.Problems associated with average alcoholconsumption and frequency of intoxicationin a medical population. Alcohol Clin EApRes. 1990;14:119-126.

21. Fiore MC, Novotny TE, Pierce JP, et al.Methods used to quit smoking in theUnited States. JAA'4. 1990;263:2760-2765.

22. Orleans CT, Schoenbach VJ, Wagner EH,et al. Self-help quit smoking interventions:effects of self-help materials, social supportinstructions, and telephone counseling. JConsult Clin Psychol. 1991;59:439-448.

23. De Soto CB, O'Donnell WE, De Soto JL.Long-term recovery in alcoholics. AlcoholClin Exp Res. 1989;13:693-697.

24. Sobell MB, Sobell LC, Kozlowski LT. Dualrecoveries from alcohol and smoking prob-lems. In: Alcohol and Tobacco: BasicScience to Policy. Rockville, Md: NationalInstitute on Alcohol Abuse and Alcohol-ism; 1995:207.

25. Pattison EM, Sobell MB, Sobell LC.Emerging Concepts ofAlcohol Dependence.New York, NY: Springer; 1977.

26. Polich JM, Kaelber CT. Sample surveysand the epidemiology of alcoholism. In:Schuckit MA, ed. Alcohol Patterns andProblems. New Brunswick, NJ: RutgersCenter of Alcohol Studies; 1985.

27. Armor DJ, MeshkoffJE. Remission amongtreated and untreated alcoholics. In: MelloNK, ed. Advances in Substance Abuse:Behavioral and Biological Research. Green-wich, Conn: JAI Press; 1983; 3.

28. Humphreys K, Moos RH, Finney JW. Twopathways out of drinking problems withoutprofessional treatment.AddictBehav. 1995;20:427-441.

29. Vaillant GE. TheNatural History ofAlcohol-ism: Causes, Patterns, and Paths to Recovery.Cambridge, Mass: Harvard UniversityPress; 1983.

30. Sobell LC, Sobell MB, Toneatto T, Leo GI.What triggers the resolution of alcohol

problems without treatment? Alcohol ClinExp Res. 1993;17:217-224.

31. Roizen R, Cahalan D, Shanks P. Spontane-ous remission among untreated problemdrinkers. In: Kandel DB, ed. LongitudinalResearch on Drug Use: Empirical Findingsand Methodological Issues. Washington,DC: Hemisphere; 1978.

32. Rosenberg H. Prediction of controlleddrinking by alcoholics and problem drink-ers. Psychological Bull. 1993;113:129-139.

33. Cunningham JA, Sobell LC, Sobell MB,Kapur G. Recovery from alcohol problemswith and without treatment: reasons forchange. J SubstAbuse. 1995;7:365-372.

34. Anthony JC, Warner LA, Kessler RC.Comparative epidemiology of dependenceon tobacco, alcohol, controlled substances,and inhalants: basic findings from theNational Comorbidity Survey. Exp ClinPsychopharmacol. 1994;2:244-268.

35. Hilton M. Drinking patterns and drinkingproblems in 1984: results from a generalpopulation survey. Alcohol Clin Exp Res.1987;11:167-175.

36. Wessinger C, Wodarski JS. Accuracy ofnational surveys of drug use: a comparisonwith local studies. J Alcohol Drug Educ.1994;39:62-73.

37. Eighth SpecialReport to the U.S. Congress onAlcohol and Health. Washington, DC:National Institute on Alcohol Abuse andAlcoholism; 1993.

............... -.a xdf| 1merstO k.

sypsao hePi-nil--NAbstracts of 250 ~....... gm t - --

972 American Journal of Public Health July 1996, Vol 86, No 7